frame rates <=> data rates
Thu, 12 Oct 1995 15:18:12 -0400

>>We are setting up cu-seeme on a wan and are getting very slow frame rates
>>although are transmission rates are very high 80-150 kbits/sec. We are using
>>the reflector program on an hp 9000. What are we doing wrong? frame rates
>>are generally 0 -1 with 8-bit encapsulated mode. Thanks for help.

For a fixed transmission rate, unusually low video frame rates could be
explained by:

1) mis-configured compression parameters
2) lots of "motion" in the video
3) competition with other data such as audio, text, or slides

The first is the most likely. For maximum frame rates, use "Standard
Resolution," set the "Change Tolerance" as high as possible without
observing corruption in the video, and set the "Refresh Interval" to at
least 100. Using "High Resolution" has no effect on video quality unless
recipients are viewing the large size 240 x 320) image, and it cuts the
frame rate dramatically. The correct value for "Change Tolerance" depends
upon the camera, digitzer and lighting conditions. You do not need to be
transmitting video to observe it's effect, but you do need to move around.
Increase the value of Change Tolerance until you start seeing parts of the
picture "tear off" as you move. Then, back it down until generation of
these artifacts is reduced to an acceptable level.

Frame rates will also be reduced the more each frame is different from the
preceding. This will occur naturally if there is a lot of movement in the
scene, but can sometimes be a spurious result of lighting effects. To
assess this possibility, connect to self, and then change the size (zoom
box) of the second "received" video image. A change in window size will
cause the video to go gray, and then to fill in, only as the changing parts
of the image are updated. Parts of the image that are supposedly static
should only fill in gradually, as the "Refresh Interval" (how many frames
are allowed to elapse before an unchanged portion of the picture is forced
to update) expires (for the purpose of this test, you might want to
increase Refresh Interval to a large value). If background is being
unnecessarily updated, you try increasing the "Change Tolerance," or try
changing the lighting or background (e.g., flourescent light cycles can
cause spurious "motion"). As an interesting example of the kind of
unexpected phenomenon that can occur, someone alerted me to the fact that
my data rate would rise dramatically at night, when my office was
completely dark. Apparently, the digitizer or camera was increasing its
sensitivity to the point that it was detecting random noise in the image
(that was introduced by some other part of the system).

3) is an unlikely possibility, as these data streams generally don't demand
a lot of bandwidth compared to video. On the other hand, audio set to the
64 kbps encoding, and being transmitted continuously, would take a pretty
big chunk out of 80-150 kbps.

The above comments are Mac-centric. The basic concepts should be identical
for Windows, though I'm not certain that all of the user interface items
are in place at this time.

Tim Dorcey
Sr. Programmer/Analyst (607) 255-5715
Advanced Technologies & Planning
CIT Network Resources
Cornell University
Ithaca, NY 14850